AI cyber incident

WSJ reports that Visa's Project Glasswing testing showed frontier AI could link minor weaknesses into viable attack chains, shifting the operational burden from discovery to validation, prioritization, and remediation. Visa developed the Visa Vulnerability Agentic Harness to automate validation, patching, and remediation testing.

What it really brings to bear is that it can construct attack chains. It will shift the emphasis from finding issues to validating, prioritizing and fixing at a level of speed and automation to stay ahead of what will be machine-speed attackers.

CISA issued Binding Operational Directive 26-04, making risk-based vulnerability remediation compulsory for federal civilian executive branch agencies. The directive explicitly says AI may further narrow defender reaction time between patch release and possible exploitation.

Cyber threat actors exploit unpatched vulnerabilities, and their use of AI may further narrow the time defenders have to react between patch release and possible exploitation. As a result, we must take immediate action to harden American networks and ensure our cybersecurity practices, including our policies for applying patches, address modern and increasingly sophisticated cyber threats.

Anthropic launched Claude Mythos 5 for a small group of cyberdefenders and infrastructure providers through Project Glasswing, in collaboration with the US government. Anthropic says the model is the same underlying model as Fable 5 but with cyber safeguards lifted in some areas and calls it the strongest cybersecurity model in the world.

For a small group of cyberdefenders and infrastructure providers, we're also launching Claude Mythos 5. It's the same underlying model as Fable 5, but with the safeguards lifted in some areas.

EVERTEC, a payment processor serving Puerto Rico financial institutions, disclosed via voluntary Item 8.01 that an unauthorized party obtained financial institution clients' transaction records and payment card data through a third-party support platform. No AI involvement identified; no operational disruption; no quantified dollar loss; filed as Item 8.01 not Item 1.05 -- does not meet material cybersecurity incident threshold.

the Company believes that an unauthorized party obtained, through a third-party support platform, certain of our financial institution clients' information related to transaction records, payment card numbers of some customers and, in some instances, customer names and contact information.

Microsoft Threat Intelligence reports that Anthropic's Claude Code GitHub Action could expose CI/CD workflow secrets when AI agents process untrusted GitHub issue, pull request, or comment content. Anthropic mitigated the issue in Claude Code version 2.1.128 by blocking sensitive /proc file access, but Microsoft says AI workflows with untrusted content and secrets should be treated as high risk.

Microsoft Threat Intelligence discovered that Anthropic's Claude Code GitHub Action could expose CI/CD workflow secrets when AI agents process untrusted GitHub content, including issue bodies, pull request descriptions, and comments. Following our responsible disclosure, Anthropic mitigated this issue in Claude Code version 2.1.128 by blocking access to sensitive /proc files.

CrowdStrike says autonomous AI agents should be treated as privileged identities because they can access systems and trigger workflows at speed and scale. The report frames agent compromise as a new attack path across enterprise environments, while offering defensive principles rather than reporting an actual systemic incident.

AI agents behave like users, but operate at a speed and scale no human can match. If compromised, an AI agent can give an adversary legitimate access to move quickly across environments, creating a new attack path that security teams can't afford to ignore.

Microsoft AI Red Team updated its agentic AI failure-mode taxonomy after 12 months of red team engagements against deployed agentic systems. The report adds seven new failure-mode categories and cites a security audit that found 512 vulnerabilities, including CVE-2026-25253, in a mainstream open-source agentic framework.

The update adds seven new failure mode categories, expands the mitigations section, and grounds the framework in 12 months of red team engagements against deployed agentic systems. A security audit conducted shortly after launch identified 512 vulnerabilities including CVE-2026-25253, a one-click RCE via WebSocket hijacking.

OpenAI published a frontier AI governance blueprint that calls for a national framework, a strengthened CAISI, and a broader government resilience plan for national security and public safety challenges from frontier AI. This is a defensive governance response, not evidence of a threshold-crossing cyber incident.

The blueprint outlines a three-part strategy: building a national framework that leverages the emerging consensus reflected in state frontier safety laws; strengthening CAISI as the U.S. federal government's primary institution for frontier AI safety; and mobilizing a broader resilience plan across government to address the national security and public safety challenges posed by frontier AI.

Anthropic says it is expanding Project Glasswing from the original roughly 50 partners to approximately 150 new organizations after coordination with partners, security industry participants, open-source maintainers, and the US government. The new group spans more than 15 countries and includes organizations providing critical infrastructure, extending controlled access to Mythos-class defensive capability while increasing operational diffusion.

Following several weeks of close collaboration with our Project Glasswing partners, the security industry, open-source software maintainers, and the US government, we're extending the partnership to approximately 150 new organizations. The organizations in this new group are based in more than 15 countries, and most provide critical infrastructure to many more.

The White House executive order directs CISA to issue binding operational directives and other guidance for AI-enabled cyber defense and directs Treasury, NSA, CISA, and the National Cyber Director to form an AI cybersecurity clearinghouse for vulnerability scanning, validation, remediation, and patch distribution. This is a new official US government action responding to frontier AI cyber risk, including access for operators such as community banks and local utilities.

Within 30 days of the date of this order, the Secretary of Homeland Security, through the Director of the Cybersecurity and Infrastructure Security Agency (CISA), in consultation with the Director of the Office of Management and Budget (OMB), the Assistant to the President for National Security Affairs, and the National Cyber Director, shall release Binding Operational Directives and other guidance as appropriate to: expedite and prioritize the cyber defense of civilian Federal Government information systems in order to protect our Nation's vital functions; establish or expand Federal programs and cybersecurity services that enhance AI-enabled defensive tools; and facilitate access to cybersecurity tools and services including, where appropriate, covered frontier models for agencies, State and local authorities, and operators of critical infrastructure such as rural hospitals, community banks, and local utilities. Within 30 days of the date of this order, the Secretary of the Treasury, in consultation with the National Cyber Director, the Secretary of War, through the Director of the National Security Agency (NSA), and the Secretary of Homeland Security, through the Director of CISA, shall form an AI cybersecurity clearinghouse, in voluntary collaboration with the AI industry and operators of critical infrastructure, that coordinates and deconflicts scanning for software vulnerabilities, discovers and validates such vulnerabilities, and coordinates and prioritizes remediation and distribution of vulnerability patches.